17,000 manually annotated Facebook comments from Tunisian media outlets categorized into positive and negative sentiment polarities. The data spans an 18-month period from January 2015 to June 2016, specifically targeting interactions on pages like Mosaique FM and Nessma TV.
Use Cases
- Train sentiment classifiers for Tunisian dialect using the manually annotated polarity labels
- Analyze public opinion trends regarding Tunisian media outlets by processing comments from Mosaique FM and Nessma TV
- Develop natural language processing tools specifically for North African Arabic dialects based on the raw comment text
Strengths
- 17,000 manually annotated user comments
- Binary sentiment labels covering positive and negative polarities
- Data sourced from 5 major Tunisian media outlets including Mosaique FM and Nessma TV
- Temporal coverage spanning 18 months from January 2015 to June 2016